understanding MCUs: Flash vs. SRAM

Translate This Thread From English to

Threaded View
Joe Strout's Avatar (by Gravatar)Joe Strout posted on
May 22, 2007, 1:48 pm
rate this thread
I've been reading a lot of microcontroller specs today, and I want to be
sure I understand what I'm reading.  Most of them have at least two
kinds of memory.  Please let me know if this is right:

Flash Memory: this is written in the course of programming the chip, but
not modified when the program is running.  Most kinds of flash memory
can be overwritten a limited (but probably very large) number of times.  
Flash Memory of course keeps its contents when power is removed.

Internal RAM (also called SRAM): this can be modified while the program
is written.  Does this also keep its contents when power is removed?

Some MCUs have, in addition to these, some EEPROM memory.  I know what
EEPROM is, but I don't know how an MCU would make use of it.  Can anyone
provide a clue?


Assuming my understanding is correct, it's disappointing how limited
microcontrollers are on internal RAM.  PIC- and AVR-based ones have SRAM
measured in bytes (or small kb, e.g. the 4kb on the ATMega128).  Even
the ET-ARM Stamp (using an LPC2119 ARM chip) has 128 kb of Flash, but
only 16 kb of internal RAM.

What I'd really like to find -- no doubt for illogical comfort reasons
-- is something like an Apple II on a chip: 16k of Flash, and 48k or
more of RAM.  In theory, you could then put Applesoft BASIC (or
something very similar) into the Flash, along with a "DOS" for accessing
say an SD card, and then have the full 48k for BASIC programs and
variables.  But I'm just dreaming...

Thanks,
- Joe

Other Related Threads
Joe Pfeiffer's Avatar (by Gravatar)Joe Pfeiffer posted on
May 22, 2007, 2:10 pm
Permalink

Re: understanding MCUs: Flash vs. SRAM



No -- SRAM "forgets" when it loses power.


I don't think I've seen one with both flash and eeprom -- regardless,
eeprom is used just like flash.  The difference between them is that
eeprom is more flexible, as you can erase a single word at a time
rather than a whole segment.  It's also more expensive.

I guess if you had both flash and eeprom you'd use flash for things
like the program, and eeprom to hold configuration parameters or logs
that have to survive over power cycles.


In general, projects that are appropriate for little 8-bit microes
don't need all that much memory (but see below).


That isn't how you'd generally do it -- the basic program would also
be in the flash or eeprom.  All you'd have in ram would be variables.
You'd use an emulator for most debugging, and burn your program into
flash or eeprom for on-robot testing.  These chips aren't really
intended for the sort of interactive use that an Apple ][ was.


Other Related Threads
Gordon McComb's Avatar (by Gravatar)Gordon McComb posted on
May 22, 2007, 4:50 pm
Permalink

Re: understanding MCUs: Flash vs. SRAM


Many MCUs use SRAM but it's not an absolute given. The data does not
remain when power is removed; it simply does not need constant
refreshing as dynamic RAM does. Remove power and this memory is lost.


Storing data values, lookup tables that may be altered during the course
of operation, etc. You write to EEPROM what you want to have remain even
when power is removed.


It's not usually a problem for most applications unless you're doing
something that requires larger chunks of memory. One case might be
interfacing to DOS FAT (for reading/writing SD cards for example). You'd
want a few K of RAM in order to keep enough data in memory to do
something meaningful. But for the most part the typical MCU program
takes up <1-2 KB.


It just invites bloated coded and there aren't many application that
would require it. However, when you need more RAM most MCU's support
connection to an external RAM chip, usually via SPI or IC2. This is a
better approach because those users who don't need an MCU with so much
RAM don't have to pay for it, and Atmel/Microchip/etc. don't have to
stock so many variations of their products. Cheaper for everybody.

-- Gordon

John Nagle's Avatar (by Gravatar)John Nagle posted on
May 22, 2007, 9:48 pm
Permalink

Re: understanding MCUs: Flash vs. SRAM



    A reasonably close approximation is an ATMEGA640-16AU, which has
64K of flash and 8K of RAM, and delivers about 16 MIPS. The part
itself costs about $12 in quantity one.

    "64-Kbyte self-programming Flash Program Memory, 8-Kbyte SRAM, 4-KByte
EEPROM, 16 Channel 10-bit A/D-converter. JTAG interface for on-chip-debug. Up to
16 MIPS throughput at 16 MHz. 1.8 - 5.5 Volt Operation."

    There's a free tool chain for the thing from Atmel (it's really GCC
plus downloaders and debuggers), and a good selection of low-cost boards.

    If you want more RAM, you can get a board with a memory socket and
plug in up to 64K.

    A useful board for learning this thing is

http://microcontrollershop.com/product_info.php?products_idX8

which costs US $54 and has a slightly differerent ATMega:
128 kBytes Program Flash, 4 kBytes data EEPROM, 4 kBytes RAM.
Less RAM, more Flash.  That board has five pushbuttons, a
2-line 16 char LCD, a LED, and a power relay, plus all the microcontroller
I/O ports and a DB-9 connector.  For $17 extra you can get a power supply
and data cable to go with it.  All the necessary software is free.  With
that and a PC, you have everything you need to get started.

                    John Nagle

Joe Strout's Avatar (by Gravatar)Joe Strout posted on
May 23, 2007, 12:11 am
Permalink

Re: understanding MCUs: Flash vs. SRAM



Thanks John, that does look like a nice one.


Yes, that looks good.  Seems like the ATmega128 is quite popular, too; I
see it pop up in a lot of different boards.  This looks like a decent
one (though I'm still rather fond of Pololu, whose Orangutan line is
similar to this in many ways).

Best,
- Joe

Other Related Threads
Alex Gibson's Avatar (by Gravatar)Alex Gibson posted on
May 26, 2007, 2:41 am
Permalink

Re: understanding MCUs: Flash vs. SRAM



My favourite avr board is the mavaric from www.bdmicro.com
Brian also has a nice hbridge board http://www.bdmicro.com/rx50/
http://www.bdmicro.com/mavric-iib/
sample code for the board
http://www.bdmicro.com/code/

Joe there is an avr gcc package in the fink and darwin ports collections but
they are a bit out of date.

Look at www.digithink.com/OSX_AVR  for up to date version of gcc, avr dude
and
other utilities for OSX. They have a nice easy to install dmg file.

or look at Mikey Sklar's instructions for using fink
screwdecaf.cx/tools.html

Alex



Joe Strout's Avatar (by Gravatar)Joe Strout posted on
May 26, 2007, 6:02 pm
Permalink

Re: understanding MCUs: Flash vs. SRAM



That sounds great!  But... where is it?  The link you mention doesn't
exist, but I did a search and found

  http://www.digithink.com/ActiveProjects/Embedded

...but the complete contents of that page, apart from links to other
pages, is:

  Embedded(ActiveProjects){
    * The flicker mechanism.
    * Building and maintaining the Avr-gcc/avra/avrdude tools for OSX.
  }

...To which I say, sounds great, but where is it?  One of the most
obtuse web pages I've ever seen.


I've used Fink in the past but wasn't very happy with it, and if the
packages therein are outdated, I'd much rather just build from source.

But if the digithink folks really do have a nice up-to-date DMG hidden
somewhere on their site, that'd be very helpful.

Best,
- Joe

Site Timeline